verify-this — Claude Code skill: prove a claim with local evidence
The skill tests a specific claim against fresh local evidence: it captures before and after measurements, compares artifacts, and returns one verdict.
npx skills add cursor/plugins --skill verify-this -a claude-codeWhat verify-this does
The skill restates a claim so it can be disproved: condition, metric, threshold. Then it picks the smallest local surface that could break the claim and measures the old and new states with the same command, data and environment. What it compares is raw output: numbers, screenshots, terminal transcripts, HTTP responses, profiles, heap snapshots.
The verdict is exactly one of VERIFIED, NOT VERIFIED or INCONCLUSIVE. A negative result gets no softening. If there is no valid baseline, the signal is noisy or a measurement fails, the skill says the comparison does not hold.
Artifacts go to /tmp/verify-this/<claim-slug>/ — claim.md, timeline.md, baseline/, treatment/, diff/ and verdict.md. Those files can carry sensitive code, prompts, screenshots or HTTP bodies, so writing them to disk is worth doing only when you agree to it; otherwise the evidence stays inline in the conversation. Vague claims like “the code is cleaner” are out of scope — the skill asks for something measurable first.
What it helps with
- prove a bug fix actually works instead of assuming it stopped reproducing
- compare behaviour before and after a change with one command
- measure a claimed performance or memory improvement
- confirm that user-visible behaviour matches a passing test
- get a clear answer instead of a recap of the diff
Example prompts for the agent
- Verify that this change made the list endpoint faster — capture before and after timings and compare
- Prove the button works again, with a screenshot of the old and new state
- Check the claim that the new query does not leak memory and give me a verdict
How to install verify-this
1Skills CLI (Vercel)
npx skills add cursor/plugins --skill verify-this -a claude-code2Manually: put the files in place
~/.claude/skills/verify-this/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/verify-this/
1Skills CLI (Vercel)
npx skills add cursor/plugins --skill verify-this -a codex2Manually: put the files in place
.agents/skills/verify-this/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add cursor/plugins --skill verify-this -a cursor2Manually: put the files in place
.cursor/skills/verify-this/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install cursor/plugins verify-this2Manually: put the files in place
.github/skills/verify-this/SKILL.md1Skills CLI (Vercel)
npx skills add cursor/plugins --skill verify-this -a gemini-cli2Platform CLI
gemini skills install https://github.com/cursor/plugins --path cursor-team-kit/skills/verify-thisManually: put the files in place
.devin/skills/verify-this/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add cursor/plugins --skill verify-this -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about verify-this
How do I install verify-this in Claude Code?
Run npx skills add cursor/plugins --skill verify-this -a claude-code. The skill lands in ~/.claude/skills/verify-this/SKILL.md.
What does verify-this do?
It restates a claim in falsifiable form, captures baseline and treatment measurements, compares the artifacts and returns VERIFIED, NOT VERIFIED or INCONCLUSIVE.
What do I need for it to work?
A measurable claim with a metric and a threshold. For vague statements the skill asks you to make it concrete first.
